ylhan/note_sync
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
Files
14c280ffee1388c74ef8347008d464d56e377418
note_sync/大模型/deepseek/README.md
T

29 KiB
Raw Blame History

DeepSeek 官方论文合集（原文 + 中文翻译）

本目录收录 深度求索（DeepSeek / DeepSeek-AI）自成立以来以论文形式公开发布的全部研究成果，共 33 篇（截至 2026-09-20）。

每篇论文均包含两部分：

  • 原文 PDF：从 arXiv 下载的官方预印本（未做改动）
  • 中文 Markdown：摘要全文翻译 + 核心贡献 + 方法要点 + 实验结果 + 结论，便于快速掌握论文内容

说明：本次中文版本为「摘要 + 核心贡献」精读版。如需某篇论文的全文逐段翻译，可在此目录基础上追加。

收录范围与判定依据

论文清单通过以下渠道交叉核验后确定，凡判定为 DeepSeek 官方成果者全部收录：

  1. DeepSeek 官网「研究索引」（https://www.deepseek.com/news ）——官方自行列出的论文条目
  2. arXiv 作者检索：作者字段含 DeepSeek-AI 的全部论文
  3. 论文署名单位核验：逐篇抓取 arXiv 摘要页，确认作者单位为 DeepSeek-AI / DeepSeek AI
  4. 社区索引交叉比对：Hugging Face DeepSeek Papers 合集、第三方综述

判定标准：论文作者单位标注为 DeepSeek-AI / DeepSeek AI，或作者主体为 DeepSeek 在职研究员。仅「以 DeepSeek 模型为研究对象」的第三方论文（数量众多）不予收录。

另有一项非论文形式的研究发布：LPLB（Linear-Programming-Based Load Balancer）——DeepSeek 于 2025 年 11 月 1 日发布的 MoE 专家并行负载均衡器，以 GitHub 仓库形式发布（早期研究阶段），未上传 arXiv 预印本，故不在本目录论文之列。

目录结构

大模型/deepseek/
├── README.md                   本索引
├── 01-基础模型/ … 09-基础设施/    各篇论文的原文 PDF + 中文 Markdown（同名配对）
├── _text/                      论文正文纯文本（供后续「全文逐段翻译」使用）
├── _digest/                    摘要 + 章节标题 + 正文首尾（精读写作素材）
└── _scripts/                   清单核验、PDF 下载、正文提取、索引生成脚本

_text/ 与 _digest/ 已按 arXiv 编号逐篇存好论文全文，后续若要把某几篇扩写为全文逐段翻译，可直接基于这些文件进行，无需重新抓取。

DreamCraft3D（2310.16818）为 DeepSeek 参与度有限的早期合作论文：作者单位列入 DeepSeek AI，论文注明「实习期间部分完成」，为完整起见一并收录并单独标注。

目录

论文总览

# 论文 arXiv 发布日期 类别
1 DeepSeek LLM: Scaling Open-Source Language Models with Longtermism 2401.02954 5 Jan 2024 基础模型
2 DeepSeekMoE: Towards Ultimate Expert Specialization in Mixture-of-Experts Language Models 2401.06066 11 Jan 2024 基础模型
3 DeepSeek-V2: A Strong, Economical, and Efficient Mixture-of-Experts Language Model 2405.04434 7 May 2024 基础模型
4 DeepSeek-V3 Technical Report 2412.19437 27 Dec 2024 基础模型
5 DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models 2512.02556 2 Dec 2025 基础模型
6 DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence 2606.19348 26 Apr 2026 基础模型
7 DeepSeek-V4.1-Flash: Pushing the Limits of KV Cache Compression 2609.19969 17 Sep 2026 基础模型
8 DeepSeek-Coder: When the Large Language Model Meets Programming -- The Rise of Code Intelligence 2401.14196 25 Jan 2024 代码模型
9 DeepSeek-Coder-V2: Breaking the Barrier of Closed-Source Models in Code Intelligence 2406.11931 17 Jun 2024 代码模型
10 CodeI/O: Condensing Reasoning Patterns via Code Input-Output Prediction 2502.07316 11 Feb 2025 代码模型
11 DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models 2402.03300 5 Feb 2024 数学模型
12 DeepSeekMath-V2: Towards Self-Verifiable Mathematical Reasoning 2511.22570 27 Nov 2025 数学模型
13 DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning 2501.12948 22 Jan 2025 推理模型
14 DeepSeek-VL: Towards Real-World Vision-Language Understanding 2403.05525 8 Mar 2024 多模态
15 Janus: Decoupling Visual Encoding for Unified Multimodal Understanding and Generation 2410.13848 17 Oct 2024 多模态
16 DeepSeek-VL2: Mixture-of-Experts Vision-Language Models for Advanced Multimodal Understanding 2412.10302 13 Dec 2024 多模态
17 Janus-Pro: Unified Multimodal Understanding and Generation with Data and Model Scaling 2501.17811 29 Jan 2025 多模态
18 DeepSeek-OCR: Contexts Optical Compression 2510.18234 21 Oct 2025 多模态
19 DeepSeek-OCR 2: Visual Causal Flow 2601.20552 28 Jan 2026 多模态
20 JanusFlow: Harmonizing Autoregression and Rectified Flow for Unified Multimodal Understanding and Generation 2411.07975 12 Nov 2024 多模态
21 DreamCraft3D: Hierarchical 3D Generation with Bootstrapped Diffusion Prior 2310.16818 25 Oct 2023 多模态
22 DeepSeek-Prover: Advancing Theorem Proving in LLMs through Large-Scale Synthetic Data 2405.14333 23 May 2024 形式化证明
23 DeepSeek-Prover-V1.5: Harnessing Proof Assistant Feedback for Reinforcement Learning and Monte-Carlo Tree Search 2408.08152 15 Aug 2024 形式化证明
24 DeepSeek-Prover-V2: Advancing Formal Mathematical Reasoning via Reinforcement Learning for Subgoal Decomposition 2504.21801 30 Apr 2025 形式化证明
25 mHC: Manifold-Constrained Hyper-Connections 2512.24880 31 Dec 2025 架构
26 Conditional Memory via Scalable Lookup: A New Axis of Sparsity for Large Language Models 2601.07372 12 Jan 2026 架构
27 Native Sparse Attention: Hardware-Aligned and Natively Trainable Sparse Attention 2502.11089 16 Feb 2025 架构
28 Auxiliary-Loss-Free Load Balancing Strategy for Mixture-of-Experts 2408.15664 28 Aug 2024 架构
29 Let the Expert Stick to His Last: Expert-Specialized Fine-Tuning for Sparse Architectural Large Language Models 2407.01906 2 Jul 2024 训练方法
30 Inference-Time Scaling for Generalist Reward Modeling 2504.02495 3 Apr 2025 训练方法
31 Fire-Flyer AI-HPC: A Cost-Effective Software-Hardware Co-Design for Deep Learning 2408.14158 26 Aug 2024 基础设施
32 Insights into DeepSeek-V3: Scaling Challenges and Reflections on Hardware for AI Architectures 2505.09343 14 May 2025 基础设施
33 DualPath: Breaking the Storage Bandwidth Bottleneck in Agentic LLM Inference 2602.21548 25 Feb 2026 基础设施

基础模型

通用大语言模型系列技术报告（LLM / MoE / V2 / V3 / V3.2 / V4 / V4.1-Flash）

DeepSeek LLM: Scaling Open-Source Language Models with Longtermism

  • arXiv：2401.02954 ｜ 发布：5 Jan 2024
  • 作者：DeepSeek-AI : Xiao Bi , Deli Chen , Guanting Chen , Shanhuang Chen , Damai Dai , Chengqi Deng , Honghui Ding , Kai Dong , Qiushi Du , Zhe Fu , Huazuo Gao , Kaige Gao , Wenjun Gao , Ruiqi Ge , Kang Guan , Daya Guo , Jianzhong Guo , Guangbo Hao , Zhewen Hao , Ying He , Wenjie Hu , Panpan Huang , Erhan…
  • 原文 PDF：01-DeepSeek-LLM-2401.02954.pdf
  • 中文翻译：01-DeepSeek-LLM-2401.02954.md

DeepSeekMoE: Towards Ultimate Expert Specialization in Mixture-of-Experts Language Models

  • arXiv：2401.06066 ｜ 发布：11 Jan 2024
  • 作者：Damai Dai , Chengqi Deng , Chenggang Zhao , R.X. Xu , Huazuo Gao , Deli Chen , Jiashi Li , Wangding Zeng , Xingkai Yu , Y. Wu , Zhenda Xie , Y.K. Li , Panpan Huang , Fuli Luo , Chong Ruan , Zhifang Sui , Wenfeng Liang
  • 原文 PDF：02-DeepSeekMoE-2401.06066.pdf
  • 中文翻译：02-DeepSeekMoE-2401.06066.md

DeepSeek-V2: A Strong, Economical, and Efficient Mixture-of-Experts Language Model

  • arXiv：2405.04434 ｜ 发布：7 May 2024
  • 作者：DeepSeek-AI , Aixin Liu , Bei Feng , Bin Wang , Bingxuan Wang , Bo Liu , Chenggang Zhao , Chengqi Dengr , Chong Ruan , Damai Dai , Daya Guo , Dejian Yang , Deli Chen , Dongjie Ji , Erhang Li , Fangyun Lin , Fuli Luo , Guangbo Hao , Guanting Chen , Guowei Li , H. Zhang , Hanwei Xu , Hao Yang , Haowei…
  • 原文 PDF：03-DeepSeek-V2-2405.04434.pdf
  • 中文翻译：03-DeepSeek-V2-2405.04434.md

DeepSeek-V3 Technical Report

  • arXiv：2412.19437 ｜ 发布：27 Dec 2024
  • 作者：DeepSeek-AI , Aixin Liu , Bei Feng , Bing Xue , Bingxuan Wang , Bochao Wu , Chengda Lu , Chenggang Zhao , Chengqi Deng , Chenyu Zhang , Chong Ruan , Damai Dai , Daya Guo , Dejian Yang , Deli Chen , Dongjie Ji , Erhang Li , Fangyun Lin , Fucong Dai , Fuli Luo , Guangbo Hao , Guanting Chen , Guowei Li…
  • 原文 PDF：04-DeepSeek-V3-2412.19437.pdf
  • 中文翻译：04-DeepSeek-V3-2412.19437.md

DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models

  • arXiv：2512.02556 ｜ 发布：2 Dec 2025
  • 作者：DeepSeek-AI , Aixin Liu , Aoxue Mei , Bangcai Lin , Bing Xue , Bingxuan Wang , Bingzheng Xu , Bochao Wu , Bowei Zhang , Chaofan Lin , Chen Dong , Chengda Lu , Chenggang Zhao , Chengqi Deng , Chenhao Xu , Chong Ruan , Damai Dai , Daya Guo , Dejian Yang , Deli Chen , Erhang Li , Fangqi Zhou , Fangyun …
  • 原文 PDF：05-DeepSeek-V3.2-2512.02556.pdf
  • 中文翻译：05-DeepSeek-V3.2-2512.02556.md

DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence

  • arXiv：2606.19348 ｜ 发布：26 Apr 2026
  • 作者：DeepSeek-AI , Anyi Xu , Bangcai Lin , Bing Xue , Bingxuan Wang , Bingzheng Xu , Bochao Wu , Bowei Zhang , Chaofan Lin , Chen Dong , Chenchen Ling , Chengda Lu , Chenggang Zhao , Chengqi Deng , Chengyu Hou , Chenhao Xu , Chenze Shao , Chong Ruan , Conner Sun , Damai Dai , Daya Guo , Dejian Yang , Del…
  • 原文 PDF：06-DeepSeek-V4-2606.19348.pdf
  • 中文翻译：06-DeepSeek-V4-2606.19348.md

DeepSeek-V4.1-Flash: Pushing the Limits of KV Cache Compression

  • arXiv：2609.19969 ｜ 发布：17 Sep 2026
  • 作者：DeepSeek-AI : Anyi Xu , B. Li , Bangcai Lin , Bing Xue , BingCheng Xian , Bingzheng Xu , Bochao Wu , Bowei Zhang , Boyi Deng , C.C. Yu , Chao Jin , Chaofan Lin , Chen Dong , Chenbing Wang , Chenfan Feng , Chengda Lu , Chenggang Zhao , Chengqi Deng , Chengyuan Zhang , Chenhao Xu , Chenqi Zhao , Chenz…
  • 原文 PDF：07-DeepSeek-V4.1-Flash-2609.19969.pdf
  • 中文翻译：07-DeepSeek-V4.1-Flash-2609.19969.md

代码模型

代码智能模型与代码驱动推理（DeepSeek-Coder / Coder-V2 / CodeI/O）

DeepSeek-Coder: When the Large Language Model Meets Programming -- The Rise of Code Intelligence

DeepSeek-Coder-V2: Breaking the Barrier of Closed-Source Models in Code Intelligence

  • arXiv：2406.11931 ｜ 发布：17 Jun 2024
  • 作者：DeepSeek-AI , Qihao Zhu , Daya Guo , Zhihong Shao , Dejian Yang , Peiyi Wang , Runxin Xu , Y. Wu , Yukun Li , Huazuo Gao , Shirong Ma , Wangding Zeng , Xiao Bi , Zihui Gu , Hanwei Xu , Damai Dai , Kai Dong , Liyue Zhang , Yishi Piao , Zhibin Gou , Zhenda Xie , Zhewen Hao , Bingxuan Wang , Junxiao So…
  • 原文 PDF：02-DeepSeek-Coder-V2-2406.11931.pdf
  • 中文翻译：02-DeepSeek-Coder-V2-2406.11931.md

CodeI/O: Condensing Reasoning Patterns via Code Input-Output Prediction

数学模型

数学推理模型（DeepSeekMath / DeepSeekMath-V2）

DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models

DeepSeekMath-V2: Towards Self-Verifiable Mathematical Reasoning

推理模型

强化学习驱动的推理模型（DeepSeek-R1）

DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning

  • arXiv：2501.12948 ｜ 发布：22 Jan 2025
  • 作者：DeepSeek-AI , Daya Guo , Dejian Yang , Haowei Zhang , Junxiao Song , Peiyi Wang , Qihao Zhu , Runxin Xu , Ruoyu Zhang , Shirong Ma , Xiao Bi , Xiaokang Zhang , Xingkai Yu , Yu Wu , Z.F. Wu , Zhibin Gou , Zhihong Shao , Zhuoshu Li , Ziyi Gao , Aixin Liu , Bing Xue , Bingxuan Wang , Bochao Wu , Bei Fe…
  • 原文 PDF：01-DeepSeek-R1-2501.12948.pdf
  • 中文翻译：01-DeepSeek-R1-2501.12948.md

多模态

视觉-语言与统一多模态模型（VL / VL2 / Janus / Janus-Pro / JanusFlow / OCR 系列 / DreamCraft3D）

DeepSeek-VL: Towards Real-World Vision-Language Understanding

  • arXiv：2403.05525 ｜ 发布：8 Mar 2024
  • 作者：Haoyu Lu , Wen Liu , Bo Zhang , Bingxuan Wang , Kai Dong , Bo Liu , Jingxiang Sun , Tongzheng Ren , Zhuoshu Li , Hao Yang , Yaofeng Sun , Chengqi Deng , Hanwei Xu , Zhenda Xie , Chong Ruan
  • 原文 PDF：01-DeepSeek-VL-2403.05525.pdf
  • 中文翻译：01-DeepSeek-VL-2403.05525.md

Janus: Decoupling Visual Encoding for Unified Multimodal Understanding and Generation

  • arXiv：2410.13848 ｜ 发布：17 Oct 2024
  • 作者：Chengyue Wu , Xiaokang Chen , Zhiyu Wu , Yiyang Ma , Xingchao Liu , Zizheng Pan , Wen Liu , Zhenda Xie , Xingkai Yu , Chong Ruan , Ping Luo
  • 原文 PDF：02-Janus-2410.13848.pdf
  • 中文翻译：02-Janus-2410.13848.md

DeepSeek-VL2: Mixture-of-Experts Vision-Language Models for Advanced Multimodal Understanding

  • arXiv：2412.10302 ｜ 发布：13 Dec 2024
  • 作者：Zhiyu Wu , Xiaokang Chen , Zizheng Pan , Xingchao Liu , Wen Liu , Damai Dai , Huazuo Gao , Yiyang Ma , Chengyue Wu , Bingxuan Wang , Zhenda Xie , Yu Wu , Kai Hu , Jiawei Wang , Yaofeng Sun , Yukun Li , Yishi Piao , Kang Guan , Aixin Liu , Xin Xie , Yuxiang You , Kai Dong , Xingkai Yu , Haowei Zhang …
  • 原文 PDF：03-DeepSeek-VL2-2412.10302.pdf
  • 中文翻译：03-DeepSeek-VL2-2412.10302.md

Janus-Pro: Unified Multimodal Understanding and Generation with Data and Model Scaling

DeepSeek-OCR: Contexts Optical Compression

DeepSeek-OCR 2: Visual Causal Flow

JanusFlow: Harmonizing Autoregression and Rectified Flow for Unified Multimodal Understanding and Generation

  • arXiv：2411.07975 ｜ 发布：12 Nov 2024
  • 作者：Yiyang Ma , Xingchao Liu , Xiaokang Chen , Wen Liu , Chengyue Wu , Zhiyu Wu , Zizheng Pan , Zhenda Xie , Haowei Zhang , Xingkai yu , Liang Zhao , Yisong Wang , Jiaying Liu , Chong Ruan
  • 原文 PDF：07-JanusFlow-2411.07975.pdf
  • 中文翻译：07-JanusFlow-2411.07975.md

DreamCraft3D: Hierarchical 3D Generation with Bootstrapped Diffusion Prior

形式化证明

定理证明与形式化数学（DeepSeek-Prover 系列）

DeepSeek-Prover: Advancing Theorem Proving in LLMs through Large-Scale Synthetic Data

DeepSeek-Prover-V1.5: Harnessing Proof Assistant Feedback for Reinforcement Learning and Monte-Carlo Tree Search

DeepSeek-Prover-V2: Advancing Formal Mathematical Reasoning via Reinforcement Learning for Subgoal Decomposition

  • arXiv：2504.21801 ｜ 发布：30 Apr 2025
  • 作者：Z.Z. Ren , Zhihong Shao , Junxiao Song , Huajian Xin , Haocheng Wang , Wanjia Zhao , Liyue Zhang , Zhe Fu , Qihao Zhu , Dejian Yang , Z.F. Wu , Zhibin Gou , Shirong Ma , Hongxuan Tang , Yuxuan Liu , Wenjun Gao , Daya Guo , Chong Ruan
  • 原文 PDF：03-DeepSeek-Prover-V2-2504.21801.pdf
  • 中文翻译：03-DeepSeek-Prover-V2-2504.21801.md

架构

基础架构创新（mHC 超连接 / Engram 条件记忆 / NSA 原生稀疏注意力 / 无辅助损失负载均衡）

mHC: Manifold-Constrained Hyper-Connections

  • arXiv：2512.24880 ｜ 发布：31 Dec 2025
  • 作者：Zhenda Xie , Yixuan Wei , Huanqi Cao , Chenggang Zhao , Chengqi Deng , Jiashi Li , Damai Dai , Huazuo Gao , Jiang Chang , Kuai Yu , Liang Zhao , Shangyan Zhou , Zhean Xu , Zhengyan Zhang , Wangding Zeng , Shengding Hu , Yuqing Wang , Jingyang Yuan , Lean Wang , Wenfeng Liang
  • 原文 PDF：01-mHC-2512.24880.pdf
  • 中文翻译：01-mHC-2512.24880.md

Conditional Memory via Scalable Lookup: A New Axis of Sparsity for Large Language Models

  • arXiv：2601.07372 ｜ 发布：12 Jan 2026
  • 作者：Xin Cheng , Rui Tian , Wangding Zeng , Damai Dai , Qinyu Chen , Bingxuan Wang , Zhenda Xie , Kezhao Huang , Xingkai Yu , Chengqi Deng , Shangyan Zhou , Chenggang Zhao , Zhewen Hao , Yukun Li , Han Zhang , Zhengyan Zhang , Yixu Wei , M.Y Xu , Huishuai Zhang , Dongyan Zhao , Wenfeng Liang
  • 原文 PDF：02-Engram-条件记忆-2601.07372.pdf
  • 中文翻译：02-Engram-条件记忆-2601.07372.md

Native Sparse Attention: Hardware-Aligned and Natively Trainable Sparse Attention

Auxiliary-Loss-Free Load Balancing Strategy for Mixture-of-Experts

训练方法

训练与对齐方法（ESFT 专家专精微调 / GRM 通用奖励模型）

Let the Expert Stick to His Last: Expert-Specialized Fine-Tuning for Sparse Architectural Large Language Models

Inference-Time Scaling for Generalist Reward Modeling

基础设施

训练与推理基础设施（Fire-Flyer AI-HPC / 硬件洞察 / DualPath 存储带宽）

Fire-Flyer AI-HPC: A Cost-Effective Software-Hardware Co-Design for Deep Learning

  • arXiv：2408.14158 ｜ 发布：26 Aug 2024
  • 作者：Wei An , Xiao Bi , Guanting Chen , Shanhuang Chen , Chengqi Deng , Honghui Ding , Kai Dong , Qiushi Du , Wenjun Gao , Kang Guan , Jianzhong Guo , Yongqiang Guo , Zhe Fu , Ying He , Panpan Huang , Jiashi Li , Wenfeng Liang , Xiaodong Liu , Xin Liu , Yiyuan Liu , Yuxuan Liu , Shanghao Lu , Xuan Lu , X…
  • 原文 PDF：01-Fire-Flyer-AI-HPC-2408.14158.pdf
  • 中文翻译：01-Fire-Flyer-AI-HPC-2408.14158.md

Insights into DeepSeek-V3: Scaling Challenges and Reflections on Hardware for AI Architectures

DualPath: Breaking the Storage Bandwidth Bottleneck in Agentic LLM Inference

  • arXiv：2602.21548 ｜ 发布：25 Feb 2026
  • 作者：Yongtong Wu , Shaoyuan Chen , Yinmin Zhong , Rilin Huang , Yixuan Tan , Wentao Zhang , Liyue Zhang , Shangyan Zhou , Yuxuan Liu , Shunfeng Zhou , Mingxing Zhang , Xin Jin , Panpan Huang
  • 原文 PDF：03-DualPath-2602.21548.pdf
  • 中文翻译：03-DualPath-2602.21548.md
Reference in New Issue View Git Blame Copy Permalink